[hacknews] new /unix

henry@utzoo.UUCP (Henry Spencer) (06/05/84)

Now running on a new /unix, with several changes.  Apart from minor
cleanup work and deletion of anachronisms, the DH software has been
changed in two ways.  First, the DH driver now raises RTS as well as
DTR when a line becomes active.  This will let us go back to using
controlled carrier on the CSRG line, among others.  Second, the DH
interrupt vectors have been moved to clear some space for the vectors
of our Avalon coprocessor board.
-- 
				Henry Spencer @ U of Toronto Zoology
				{allegra,ihnp4,linus,decvax}!utzoo!henry

henry@utzoo.UUCP (Henry Spencer) (10/10/84)

Now running a new /unix with several minor improvements.  A directory with
the sticky bit set (e.g. /tmp) now does not permit removal of files that
you don't own.  Pending alarms are cancelled when execing a setuid or
setgid program.  Setuid(geteuid()) and its gid counterpart work.  And a
truly revolting kludge has been added to the tty(4) driver to permit
experimenting with the HP laser printer.
-- 
				Henry Spencer @ U of Toronto Zoology
				{allegra,ihnp4,linus,decvax}!utzoo!henry

henry@utzoo.UUCP (Henry Spencer) (11/28/84)

Now running a new /unix.  The magtape driver's error handling has been
improved in some small ways, including a reduced number of retries (going
past a couple of retries is the kiss of death anyway).  More important, the
Unibus-map allocation code has been totally rewritten to permit multiple
simultaneous allocations.  This is the major kernel mod needed to get disk
and tape running simultaneously during backups, for an expected factor of 2
performance improvement.
-- 
				Henry Spencer @ U of Toronto Zoology
				{allegra,ihnp4,linus,decvax}!utzoo!henry

henry@utzoo.UUCP (Henry Spencer) (03/05/85)

Now running on a new /unix, with a number of modifications.

The boot board from the 45 is now on the 44, which means that (at long
last!) we can dispense with lengthy typed-in boots for the disk and
tape.  Revised boot instructions will be posted shortly; the old ones
still work in a pinch.  The board is at 771000-771777.

The kernel memory-allocation stuff has been altered to deal properly
with >2MB of memory, which will probably be a reality here soon.  We
could test it right now, actually -- we have a total of 2.25MB of 44
memory on hand, although we're running with a bit under 2 right now --
but the gain doesn't seem worth fiddling with the boards, especially
when we'll probably have to do it again soon for another 1MB board.

The DH driver has been modified to permit running lines as both dialin
and dialout.  This stuff doesn't work.  We don't know why yet, although
I have a kernel dump which should tell us after a bit of analysis.  For
the moment, 5121 is still dialout-only via Wern's breakout box.

A new "device" driver has been installed:  /dev/prf is the user interface
to a kernel profiler, somewhat similar to the one in System V.  It seems
to work, although we'll need to be up for a while before it really tells
us anything useful.  Documentation is coming.

We have also done some minor cleanup chores, and removed the Star Ring
driver since Star Ring interfacing is being left to the 05s.
-- 
				Henry Spencer @ U of Toronto Zoology
				{allegra,ihnp4,linus,decvax}!utzoo!henry

henry@utzoo.UUCP (Henry Spencer) (04/30/85)

Now running a new /unix, with two minor changes.  First, the profiler
is now running off the priority-7 programmable clock, for more accurate
results.  Second, the DHs and their vectors have been moved down in
the address space; this is an "invisible" reconfiguration that should
not have any user-visible effects.  (It was done partly because of
some suspicion that the rather high addresses might have had something
to do with the problems installing the 5th DH, and partly because the
Avalon board's vector is hardwired at 500 and the 5th DH was colliding
with it.)
-- 
				Henry Spencer @ U of Toronto Zoology
				{allegra,ihnp4,linus,decvax}!utzoo!henry

henry@utzoo.UUCP (Henry Spencer) (05/28/85)

Now running on a new /unix, with an improved version of the memory-disk
driver, and with minor modifications to permit pipes to reside on the
memory-disk driver.  The result should be a general performance boost,
since we use pipes a lot.
-- 
				Henry Spencer @ U of Toronto Zoology
				{allegra,ihnp4,linus,decvax}!utzoo!henry